Cancer Risk of Patients With Overweight and Obesity: A Predictive Model

Obesity is the most common chronic disease in the U.S. Patients with obesity have many risk factors for cancer, often modifiable. It is important to identify patients with obesity at high risk of cancer to be able to appropriately direct treatment and resources. Given that the risk pool is large, it is imperative to identify a clinically meaningful metric for risk stratification to help guide interventions. We conducted an observational study of electronic medical records data for 394,161 adults aged between 18 and 80, with BMI ≥ 25 kg/m(2) and without baseline history of cancer between 2000 and 2019. We first identified a literature-based pool of risk factors for cancer onset and conducted variable selection by applying least absolute shrinkage and selection operator (LASSO) penalized Cox regression with ten-fold cross-validation on an 80% training dataset. Effects of the selected variables on risk of cancer (excluding non-melanoma skin cancer) onset were assessed using Cox regression on the 80% training dataset. The resulting model accuracy was evaluated using Cox regression on a withheld 20% validation dataset. Participants had a mean age of 46.7 (SD: 15.5) years and mean body mass index (BMI) of 30.5 (SD: 5.4) kg/m(2); 51.9% were women. Over a mean of 7.5 years of follow-up, 34,679 (8.8%) of study patients developed cancer. The predictive model achieved a Harrell’s C-statistic of 0.73. The greatest risk of cancer incidence was associated with HIV infection (HR 2.22; 95% CI 1.88–2.63; 0.27% of patients), older age (HR 2.05 per 1 SD = 15.5 years; 95% CI 2.01- 2.09), hepatitis C infection (HR 1.48; 95% CI 1.34–1.63; 0.96% of patients), and family history of cancer (HR 1.44; 95% CI 1.41–1.48; 42.5% of patients). Additional patient characteristics found in >5% of patients that also carried risk included proteinuria (5.8% of patients; HR 1.23; 95% CI 1.18–1.29) and history of smoking (40.7% of patients; HR 1.20; 95% CI 1.17–1.23). Each standard deviation increase in BMI (5.4 kg/m(2)) was associated with a hazard ratio of 1.06 (95% CI 1.05–1.07) for incident cancer. It is feasible to use predictive modeling to identify patients with overweight and obesity at high cancer risk. This approach could be utilized to guide population management and clinical treatment decisions.